Since the first current-mode control ICs emerged in the early 1980s, the popularity of current-mode control has made it the method of choice for most power supplies. Although there are a number of different types of current-mode control (Ref 1), the most popular is constant frequency with turn on at clock time. Constant-frequency control has become synonymous with current-mode control for most designers. As a result, the most popular control ICs are the 384X series, which a variety of suppliers manufacture.

Although the limitations of constant-frequency control are well known, it has remained the dominant control method for just about 10 years. Other methods, such as resonant-mode control, have not become as popular as expected, perhaps because of the difficulty in designing a practical supply and an overwhelming fear of variable switching frequencies within the power community. Industry focus recently has shifted to power-factor correction because of the impending implementation of IEC 555 and to synchronous switching for maximum efficiency. Still other methods of current-mode control remain unexplored.

One of the alternate types of current-mode control for which control ICs are available is hysteretic-mode control. Hysteretic current-mode control, which the Bose Corp patented in 1984 (Ref 2), has remained an obscure technique few designers use, yet it offers significant advantages for many applications. Hysteretic current-mode control (HCMC) offers the tightest and most accurate control of inductor current, is unconditionally stable regardless of duty cycle, and offers excellent transient response to step loads. The advantages of this control technique include

• Inherent load-current limiting
• Short-circuit-proof operation in buck-derived topologies
• No slope compensation required for duty cycles above 50%
• Easy loop-stability design
• No subharmonic oscillation
• Instantaneous response to load-current changes
• Constant peak-to-average inductor-current ratio.

You can use HCMC with most switching-regulator topologies, including buck, forward-mode, boost, and continuous-mode flyback converters (see box, "Buck-regulator design example," which takes you through a typical design step by step).

HCMC ideally suits applications that require control of both load current and output voltage and require that power supplies behave as constant-current sources and as constant-voltage regulators. Examples of these applications include battery chargers, arc welders, fluorescent lamps, laser power supplies, and servo-motor control circuits.

Due to its free-running operation, HCMC is unsuitable for any application that needs to synchronize the supply's switching frequency with some external clock. This frequency variation, while predictable, may also limit HCMC's application in certain areas, such as video. Although not discussed in this article, HCMC can provide pulse-by-pulse overcurrent protection in universal-input power supplies that use a continuous-mode flyback topology.

Compare constant frequency vs hysteretic

Comparing the operation of constant-frequency and hysteretic-mode controllers highlights major differences. Fig 1 shows a simplified block diagram of a constant-frequency current-mode regulator. The clock or oscillator sets the RS flip-flop each cycle and turns on power-switch Q₁ while the control loop determines the period for which it remains on. The external R_T/C_T network determines the operating frequency.

In Fig 1, the controller detects the peak inductor current by sampling the voltage across RS while sampling the output voltage V_OUT directly. Depending on the value of V_OUT, the output of the error amplifier determines the peak current that flows in the inductor by constantly adjusting the voltage level on the inverting terminal of the current-sense comparator. The controller detects any change in the input voltage by detecting a change in the peak current measured as a voltage across RS; it then adjusts the on-time of the FET to hold V_OUT constant.

This common configuration has several disadvantages. First, regulation performance is limited because the switch turn-on is always controlled by the clock and is independent of the feedback loop. Thus, a minimum on-time exists, which limits the maximum operating frequency and can become a significant part of the total switching period.

A second problem with this configuration also occurs: Because the circuit only detects the peak current, the peak-to-average current error will vary with duty cycle. Fig 2 shows what happens in a fixed-frequency controller when an increase in input voltage causes a change in the duty cycle (assuming a constant output voltage). Time, t₁ stands for the on and off times that correspond to some low input voltage, and t₂ to some higher input voltage. The down slope of the inductor current is constant and equal to V_OUT/L.

When V_IN is low, more time is required for the current to reach its peak value I_PEAK, which results in a higher average current (I_AVG1 in Fig 2). On the other hand, if the input voltage is high, the on-time is reduced, and the average current, I_AVG2, is lower. Thus, the peak-to-average current ratio depends on the duty cycle. Because the output voltage is proportional to the average current, changes in the input voltage cause momentary changes in the output voltage—which the error amplifier feedback loop corrects. The current-sense-amplifier loop monitors the peak current and causes a further output-voltage change.

A third problem with a fixed-frequency system is that it can become unstable for duty cycles greater than 50%. Any increase in inductor current ([delta]I₁ in Fig 3) tends to increase with time if the duty cycle is greater than 50%, resulting in a larger increase, [delta]I₂. Slope compensation can overcome this problem but adds complexity and external components to the design.

Hysteretic current-mode control

Fig 4a shows a similar regulator that uses a hysteretic-mode control system. In HCMC, no oscillator exists. The regulator senses inductor current by monitoring the voltage across R_S using a differential current-sense amplifier. This amplifier's output drives two comparators, I_PEAK and I_VALLEY. The inductor current IL ramps alternately between an upper limit I_PEAK and a lower limit I_VALLEY (Fig 4b).

The hysteretic-mode regulator (typically implemented in a control IC such as the CS324) maintains a controlled difference [delta]I between the comparators' inputs using a voltage-controlled current source (VCCS). The output of an error amplifier adjusts the reference voltage on the inverting terminal of the I_PEAK comparator depending on the value of V_OUT. The peak and valley comparators' output controls the RS flip-flop, which turns the transistor switch off and on. If V_OUT changes, the reference voltage for I_PEAK changes and the reference point for I_VALLEY follows, thereby keeping the current-hysteresis band [delta]I constant.

As the inductor current increases, the output of the current-sense amplifier reaches the threshold of comparator I_PEAK whose output goes high, which resets the flip-flop and turns off Q₁. As the inductor current ramps down, the output of the current-sense amplifier decreases to the threshold of comparator I_VALLEY. The output of this comparator then goes high, the flip-flop sets, and Q₁ turns back on.

Fig 5shows what happens when the input voltage to an HCMC supply increases. As in Fig 2, t₁ and t₂ correspond to low and high input voltages. When the input voltage is high, the switch-on time (t_2(ON)) decreases, which is similar to fixed-frequency control. However, the hysteresis band [delta]I remains constant, as does the discharge slope, which ultimately increases the switching frequency, because f₂=1/t₂. Thus, unlike constant-frequency control, the peak-to-average current difference does not vary with duty cycle.

Hysteretic control provides stability

HCMC also overcomes the instability problem associated with fixed-frequency operation for duty cycles above 50% without the need for slope compensation. An increase in load current [delta]I₁ causes the fixed hysteresis band [delta]I to move upward by an amount [delta]I₁ (Fig 6), which means that the disturbance at the end of the cycle ([delta]I₂) is the same as at the beginning. Thus, HCMC remains stable regardless of duty cycle.

Another benefit of HCMC is its ability to limit the short-circuit current in buck or buck-boost converters. When the output of a buck regulator (Fig 7) short circuits and Q₁ is on, V_OUT=0V, and the voltage across the inductor equals V_IN. The inductor current rises quickly and reduces t_ON (Fig 8). The discharge time t_OFF also increases because it is determined by V_OUT/L, and V_OUT is effectively a short circuit. The only voltage across the inductor during t_OFF is the forward voltage drop V_D of the diode. Q₁ doesn't turn on again until the current has dropped to the required valley limit. Thus, under short-circuit conditions, an HCMC regulator can only deliver a fixed maximum current.

Buck-regulator design example The design starts with a known set of requirements. For this example, VIN=9 to 14V dc, VOUT=5V dc, ±1%, IOUTMAX=2A, IOUTMIN=0.2A (10% of full load), and switching frequency=200 kHz at maximum input voltage. (1) Choose a value for the hysteresis current DI based on the minimum load conditions using Eq 12 (from text): (2) Calculate the value of RSENSE using Eq 14: (3) Choose the inductor value for continuous operation at minimum load using Eq 11. Allowing a 5% tolerance on the inductor value, LCRITICAL=43 µH ±5%. (4) Calculate the hysteresis voltage VDI using Eq 14 and Eq 15 (5) Determine minimum and maximum frequency of operation using Eq 11 (6) Design the gate-drive circuit. An N-channel FET is a better choice for the high-side switch because it has a lower RDS(ON), is less expensive, and has a wider selection available than for a comparable p-channel FET. However, the catch with using an n-channel device as a high-side switch is that it requires a gate-drive voltage. The circuit can easily generate this voltage if you add an additional winding to the main inductor. Because one end of the secondary winding is connected to the input voltage and the other to VCC, the secondary voltage will remain above VIN, providing a reliable gate-drive voltage. The steady-state voltage across the secondary winding is where VD(PR) is the voltage drop across the primary diode and VD(SEC) is the voltage drop across the secondary diode. The turns ratio in this design is 1:1, and VD(PR) and VD(SEC) are equal, so VGS=5V. (7) Calculate the maximum power dissipation in the diode and FET. If you short the output during the on-time, the regulator applies almost all of the input voltage to the inductor. However, a short during the off-time only applies the forward-diode-drop voltage to the inductor. The design must fulfill the familiar +DI=2DI balance requirement. Thus, You can use these values of short-circuit on and off times to calculate the worst-case power dissipation and power rating for the diode and FET during short-circuit conditions.

Derive design equations

In Fig 7's basic buck regulator, Q₁ interrupts the input voltage and provides a variable-duty-cycle and variable-frequency square wave to the output LC filter. The filter averages the square waves to produce a dc output voltage. The [delta]I hysteresis-band setting controls the inductor current; the voltage-feedback loop controls the output voltage. The average voltage across the inductor over one cycle is zero. The volt-time product of t_ON must equal that of t_OFF.

A number of equations are crucial to the design of a hysteretic-mode regulator. During the on and off times, the voltage across the inductor L is given by the following equations, respectively, where V_DS is the FET's drain-to-source voltage and V_D is the diode's forward voltage drop:

In a hysteretic buck regulator, the fixed hysteresis band [delta]I controls the ac current through the inductor. If [delta]I₊ is the up slope and [delta]I_- is the down slope of the inductor current in Fig 4b, then

During the on-time,

and, during the off-time,

Solving for t_ON and t_OFF produces the following equations:

Because the total cycle time T=t_ON+t_OFF and switching frequency f=1/T,

Finally, because V_D and V_DS are small compared with the other parameters in this equation, you can drop them and simplify the expression, ultimately leaving

Thus, under steady state conditions, the switching frequency is a function of V_IN alone because [delta]I, L, and V_OUT are constant.

Solve for hysteresis-band current

The average-inductor current I_AVG has a fixed relationship to the peak-inductor current I_PEAK regardless of duty cycle, and I_AVG is related linearly to the hysteresis voltage V_[delta]I. The hysteresis voltage determines the width of the current hysteresis band in Fig 4b, which, along with the inductor, determines the boundary condition between continuous and discontinuous modes of operation and thus the minimum load conditions.

A good approximation for [delta]I, to allow for some headroom before discontinuous operation begins, is

The relationship between the maximum output current I_OUTMAX and the peak inductor current I_PEAK (note that I_OUTMAX is the same quantity as I_AVG in Fig 4b) is

Sense resistor RS, which connects across the inputs to the current-sense amplifier, determines the maximum current the regulator can deliver. The relationship between R_S and the maximum voltage developed across this resistor, V_SENSEMAX, is

V_SENSEMAX has both a steady-state component due to I_OUTMAX and a varying component due to [delta]I, which is determinedby the hysteresis voltage V_[delta]I. This varying component of V_SENSEMAX is related to V_[delta]I depending on the particular IC control you use. In the case of the CS324 control IC, V_SENSE is related to [delta]I (pin 8) by the relationship:

1. Redl, Richard and Nathan O Sokal, "What a Design Engineer Should Know About Current-Mode Control," Proceedings of the Power Electronics Specialists Conference, 1985.

2. Froeschle, Thomas A, "Current-Mode Controlled Two-State Modulation," US Patent #4,456,872, January 26, 1984.

3. Neufeld, Herman, "Advantages of Variable Frequency Operation Using the 1-MHz CS320 Current Mode Controller," CSC Application Note CS037AN, Cherry Semiconductor Corp.

4. Levin, Gedaly and Kieran O'Malley, "Hysteretic current-mode control," CSC Application Note CS322/324AN, Cherry Semiconductor Corp.

5. Redl, Richard and Nathan O Sokal, "Near-Optimum Dynamic Regulation of DC-DC Converters Using Feed-Forward to Output Current and Input Voltage with Current-Mode Control," IEEE Transactions on Power Electronics, Volume PE-1, No. 3, July 1986.

6. Redl, Richard and Nathan O Sokal, "Frequency Stabilization and Synchronization of Free Running Current-Mode-Controlled Converters," Proceedings of the Power Electronics Specialists Conference, 1986.

The authors are grateful to Bob Kent of Cherry Semiconductor Corp who compiled the schematics for this article.